The AI Act’s general-purpose rules became operational
On 2 August 2025, the European Union’s obligations for general-purpose AI models entered application. The change mattered because it moved the debate from principles to duties: documentation, policies, cooperation and, for systemic-risk models, additional safeguards.
Regulation does not make a model safe by declaration. It creates obligations around the organizations that build and place models into downstream systems. The practical effect is to make provenance, risk assessment and accountability part of the infrastructure of deployment.
The AI Act’s date should therefore be read as an operational milestone, not a finish line. The hard work is the interpretation of broad rules across fast-changing model architectures and products. Policy becomes real when someone can identify the provider, the risk, the evidence and the person responsible for the decision.
